TRANSFERABLE GEOBIA DEVELOPMENT TO BUILDING IDENTIFICATION IN MULTI CHARACTERISTICS RESIDENTIAL AREA USING AERIAL PHOTOGRAPHY

• Main Author: RATNA SARI - NIM : 25115009 , LUTHFIYA
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/22943
• Institution: Institut Teknologi Bandung

Residential area diversity is a problem in building information extraction and became more challenging when associated with it's extensive space. Using 5cm spatial resolution RGB aerial photographs and nDSM height data support in 25cm spatial resolution, transferable GEOBIA developed to overcome the challenges and problems. Segmentation and classification using multilevel technique adapted in master ruleset to achieve best results with the lightest processing load. Master rulesets arranged in the main area then applied to several test areas to compliment the characteristics as well as the stability of accuracy that can be produced. Master ruleset proved to provide accurate information of reliable building area on various characteristics of residential area. In the main area, the accuracy value reaches 91.25% while in other test areas the accuracy is 68,98%, 72,80%, 78,57%, 79,95%, 86,19%, 92,22% and 92,73%. Flexibility test simulated by changing spatial resolution of RGB areal image in the range 10cm up to 150cm. With 25cm spatial resolution of height data, accuracy change always less than 5%. <br />